CN113657678A - Power grid power data prediction method based on information freshness - Google Patents

Power grid power data prediction method based on information freshness Download PDF

Info

Publication number
CN113657678A
CN113657678A CN202110970503.1A CN202110970503A CN113657678A CN 113657678 A CN113657678 A CN 113657678A CN 202110970503 A CN202110970503 A CN 202110970503A CN 113657678 A CN113657678 A CN 113657678A
Authority
CN
China
Prior art keywords
data
load
power data
power
freshness
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110970503.1A
Other languages
Chinese (zh)
Inventor
王鑫
李新民
郑国强
汪玉
秦丹丹
王峰
刘丽
张淑娟
魏李莉
赵亮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
State Grid Anhui Electric Power Co Ltd
Southwest University of Science and Technology
Original Assignee
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
State Grid Anhui Electric Power Co Ltd
Southwest University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd, State Grid Anhui Electric Power Co Ltd, Southwest University of Science and Technology filed Critical Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Priority to CN202110970503.1A priority Critical patent/CN113657678A/en
Publication of CN113657678A publication Critical patent/CN113657678A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Marketing (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Power Engineering (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Probability & Statistics with Applications (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a power grid electric power data prediction method based on information freshness, which comprises the steps of selecting an electric power data sample, selecting an elastic BP neural network to process the electric power data sample, and training in the neural network to obtain a neural network model for predicting load; step one, preprocessing electric power data; processing and sampling the power data samples; step two, calculating the freshness of the information; performing final classification on the primarily classified power data samples in the step one to obtain input power data in an output prediction stage; step three, data prediction; and (4) taking the power data samples classified in the step two as input quantity of a neural network model for predicting the load, obtaining power prediction load values of load data with different data freshness to points to be predicted, and selecting a predicted value with the minimum error as a final load predicted value. The method has higher prediction accuracy and can provide direct data support for power generation and power utilization strategies in the power system.

Description

Power grid power data prediction method based on information freshness
Technical Field
The invention relates to the field of power grid data processing, in particular to a power grid power data prediction method based on information freshness.
Background
With the development of economy and society, electric power systems are widely applied to various aspects of life and production of society. In a large enterprise, a power system is provided with a main substation and a plurality of distribution substations connected with the main substation respectively, and the main substation and one distribution substation form a loop after being connected. Therefore, the operation state of each distribution substation needs to be acquired in time, and equipment on each loop is adjusted according to the operation state of each loop.
In the prior art, power data are predicted mainly from power system influence factors so as to correspondingly adjust working conditions of power grid equipment and the like.
Disclosure of Invention
The technical problem to be solved by the invention is to solve the problem that the existing power data analysis method ignores the influence of multi-scale time sequence characteristics on the load prediction accuracy, describes the mathematical and physical relationship between multi-scale time information and prediction data from the aspect of the information freshness of the power data, and adopts a robust neural network to predict the power data, thereby providing a power grid data prediction method based on the information freshness. The method has higher prediction accuracy and can provide direct data support for power generation and power utilization strategies in the power system.
A power grid power data prediction method based on information freshness comprises a power data preprocessing stage, an information freshness calculation stage, a load prediction neural network training stage and a data prediction stage;
selecting a power data sample, processing the power data sample by using an elastic BP neural network, and training in the neural network to obtain a neural network model for predicting load;
step one, preprocessing of electric power data
Processing and sampling the power data samples, and primarily classifying the processed and sampled power data samples;
step two, calculating the freshness of the information
Performing final classification on the primarily classified power data samples in the step one to obtain input power data in an output prediction stage;
step three, data prediction
And (4) taking the power data samples classified in the step two as input quantity of a neural network model for load prediction, obtaining power prediction load values of load data with different data freshness to points to be predicted, selecting a predicted value with the minimum error as a final load predicted value, and outputting the corresponding data freshness.
In a preferred embodiment of the present inventionIn an embodiment, the sampling in the first step specifically includes: sampling the power data sample in the first step once every m minutes, wherein the sampling times are set as t, and N times of sampling are total, so that the sample set is as follows: d { (x)0,y0),(x1,y1),...,(xt,yt),...,(xN,yN) X, abscissatMt is the length of time from the point to be predicted, in seconds, and the ordinate ytPoint at time t ═ 0 (x) in kwh0,y0) As the point to be predicted for the electrical load.
In a preferred embodiment of the present invention, the primary classification in the first step specifically includes:
dividing the sample set D into n parts, and using symbol D for each sub-data setiIs expressed, and the length N of the sub data seti，
Figure BDA0003225465580000021
For each subset of power data DiMean clustering with k-means to form the best classification.
In a preferred embodiment of the present invention, the clustering algorithm formula is:
Figure BDA0003225465580000031
order (x)j,yj) Is the jth sampling result, (a)i,bi) Is the ith data centroid, then
Figure BDA0003225465580000032
Representing the euclidean distance between the jth sample and the ith centroid. By comparing the distances between the sampling point and the r centroids, the minimum distance is determined
Figure BDA0003225465580000033
And (6) carrying out data classification.
In a preferred embodiment of the present invention, the freshness of information is dependent on the subdata set DiR centroids to t 0The weighted average of the points to be predicted is defined as follows:
Figure BDA0003225465580000034
wherein (a) in the formulai,bi) Representative data subset DiThe ith centroid of (a)0,b0) I.e. the point to be predicted
Will phiiAnd subdata set DiR centroids of the load values y on the ordinateiMultiplication, the result of which is XiIs represented by Xi＝{Φiy1,Φiy2,...,Φiyr}，XiIs a column vector of r rows; mixing XiEach element X in (1)jThe Min-Max normalization of the following equation (2) is performed and then used as the input for the data prediction stage
Figure BDA0003225465580000035
In a preferred embodiment of the invention, the power data samples classified in the second step are used as input quantity of a neural network model for load prediction, power prediction load values of load data with different data freshness to points to be predicted are obtained, a predicted value with the minimum error is selected as a final load predicted value, and corresponding data freshness is output.
In a preferred embodiment of the present invention, the training process of the load prediction model is specifically described as follows:
1) information forward transfer process: output value O of input layer1And input value O of output layerLAll equal to the input value, hiding the input value x of each layer network in the layertAre the weight of the network of the layer and the output value O of the network of the previous layert-1The activation output value of the weighted sum of products.
If the information is transmitted forward from the first layer hidden layer to the t layer hidden layer, the process is as shown in formula (3):
Figure BDA0003225465580000041
for the hidden layer activation function, a sigmoid function is mainly adopted, and a specific calculation formula is shown as a formula (4):
Figure BDA0003225465580000042
2) the error back propagation process: in the process of backward transmission, firstly, the output value (namely the load predicted value) and the actual value Y of the final output layer obtained by forward transmission of the information are calculatedtruRoot mean square error E ofLAs shown in formula (5):
Figure BDA0003225465580000043
error signal ELAs a basis for modifying its weight, offset, i.e. error ELThe neurons are allocated to each level as shown in the following formula (6):
Figure BDA0003225465580000044
drawings
FIG. 1a is a flow chart of the present invention for the specific collection of a sample set; FIG. 1b is a subdata set D of the present inventioniSchematic diagram of centroid assignment.
FIG. 2 is a flow chart of the clustering algorithm of the present invention.
FIG. 3 is a flow chart of data prediction based on neural networks according to the present invention.
Detailed Description
A power grid power data prediction method based on information freshness comprises a power data preprocessing stage, an information freshness calculation stage, a load prediction neural network training stage and a data prediction stage;
selecting a power data sample, processing the power data sample by using an elastic BP neural network, and training in the neural network to obtain a neural network model for predicting load;
step one, preprocessing of electric power data
Processing and sampling the power data samples, and primarily classifying the processed and sampled power data samples;
step two, calculating the freshness of the information
Performing final classification on the primarily classified power data samples in the step one to obtain input power data in an output prediction stage;
step three, data prediction
And (4) taking the power data samples classified in the step two as input quantity of a neural network model for load prediction, obtaining power prediction load values of load data with different data freshness to points to be predicted, selecting a predicted value with the minimum error as a final load predicted value, and outputting the corresponding data freshness.
Example (b):
referring to fig. 1-3, a grid power data prediction method based on information freshness includes a power data preprocessing stage, an information freshness calculation stage, a load prediction neural network training stage and a data prediction stage;
the preprocessing stage of the power data is as follows:
selecting typical users (including but not limited to enterprises and factories with high power demand) in a certain area, sampling the power load data of the users, (for example, journal or abnormal data appears, data storage and identification are carried out, if the abnormal data is identified, the data is not taken as a prediction data set), sampling once every m minutes, setting the sampling frequency as t, and if the abnormal data is identified, sampling for N times in total, wherein the sample set is as follows: d { (x)0,y0),(x1,y1),...,(xt,yt),...,(xN,yN) X, abscissatMt is the length of time from the point to be predicted, in seconds, and the ordinate ytPoint at time t ═ 0 (x) in kwh0,y0) As the point to be predicted for the electrical load.
As shown in FIG. 1a, the sample set D is further divided into n sub-sets, each sub-set being denoted by the symbol DiIs expressed, and the length N of the sub data seti，
Figure BDA0003225465580000061
(the equal division is fair classification of data without prior information, and the data can also be divided into N parts, and each part of the data set is N in lengthi) (ii) a As shown in FIG. 1b, each sub-dataset DiA total of r centroids, (a)i,bi) Respectively, the abscissa (sampling time) and the ordinate (load value at the sampling time) of the ith centroid within the sub data set.
The result of the preliminary partitioning of the sample set is shown in fig. 2 below.
For each subset of power data DiMean value clustering is carried out by adopting k-means, and the convergence value of a clustering algorithm is defined as beta (0)<β<1)。
First in the power data subset DiAnd randomly selecting r centroids as initial centroids. Wherein the ith centroid is ziI is more than or equal to 1 and less than or equal to r, namely each power data subset DiAnd is divided into r clusters. By pair of the subsets DiTraversing the layers of all the data points, and selecting the mass center z with the minimum distance from the mass center ziAnd classifying the data into the ith cluster so as to make the subdata set DiR optimal sets of the current batch can be obtained, and the optimal means that the distance difference between any two data points in each set is smaller than the convergence value beta.
After distributing all the points of the optimal set to the centroid of the current batch, continuing the hierarchical traversal of all the data points in the r clusters in the next batch, and calculating the mean value of the internal data points of each cluster to serve as a new centroid
Figure BDA0003225465580000062
Figure BDA0003225465580000063
For convenience of description, α and β denote new centroids, respectively
Figure BDA0003225465580000064
With the initial centre of mass ZiAnd an algorithm convergence threshold value,
Figure BDA0003225465580000065
if α is<β, illustrating the power data subset DiThe loop exits if r clusters have reached the best classification.
The flow chart of the clustering algorithm is shown in fig. 1b below.
Order (a)i,bi) Is the ith data centroid, (x)j,yj) Is the jth sampling result, then
Figure BDA0003225465580000071
The distance between the sampling point j and the ith individual mass center is calculated according to an Euclidean distance formula, and the minimum distance is adopted as the classification of the data set to which the sampling point belongs through distance comparison.
Step two, the information freshness calculation stage process is as follows:
because the power data are discrete data sets in a time dimension, power data analysis is closely related to time, however, the existing data analysis method mainly starts from the front and back time sequence characteristics of the power system, and does not consider the influence of the freshness of the data on the prediction accuracy of the power data. The smaller the time difference between the sample data and the predicted data is, the fresher the sample data is, and conversely, the larger the time difference is, the more the sample data is out of date.
For each power data subset DiThe information freshness is dependent on the sub data set DiThe weighted average of r centroids to the point to be predicted where t is 0 is defined as follows:
Figure BDA0003225465580000072
wherein (a) in the formulai,bi) Representative data subset DiThe ith centroid of (a)0,b0) I.e. the point to be predicted. Tong (Chinese character of 'tong')Following the calculation of equation (1), each subset of data DiAll have corresponding unique data freshness, andithe smaller the load data information is, the fresher the load data information is. Will phiiAnd subdata set DiR centroids of the load values y on the ordinateiMultiplication, the result of which is XiIs represented by Xi＝{Φiy1,Φiy2,...,Φiyr}，XiIs a column vector of r rows. Mixing XiEach element X in (1)jAfter Min-Max normalization as in equation (2) below, the data is used as input for the data prediction stage in step four.
Figure BDA0003225465580000073
Step three, load prediction neural network training stage
And after the Min-Max normalization is carried out on the sample D which is not divided by the elastic BP neural network, the sample D is thrown into the neural network for training, and a stable neural network model which is suitable for load prediction is obtained through the self-learning capability of the neural network. The network structure is shown in fig. 3, which is sequentially an input layer (black), a hidden layer (blue) and an output layer (red) from left to right.
The training process is mainly divided into the forward transfer of information and the reverse correction process of errors, as follows:
the elastic BP neural network training process mainly comprises the forward transmission of information and the backward propagation of errors, the signs of all hyper-parameters of a specified network are as follows, the learning rate is eta, the total number of network layers is L, the number of each network neuron is PlThe weighting parameter of the l-th network is wlBias parameter blAnd O for output valuelAnd (4) showing.
The training process of the load prediction model is specifically described as follows:
1) information forward transfer process: output value O of input layer1And input value O of output layerLAll equal to the input value, hiding the input value x of each layer network in the layertAre the weight of the network of the layer and the output value O of the network of the previous layert-1The activation output value of the weighted sum of products.
If the information is transmitted forward from the first layer hidden layer to the t layer hidden layer, the process is as shown in formula (3):
Figure BDA0003225465580000081
for the hidden layer activation function, a sigmoid function is mainly adopted, and a specific calculation formula is shown as a formula (4):
Figure BDA0003225465580000082
2) the error back propagation process: in the process of backward transmission, firstly, the output value (namely the load predicted value) and the actual value Y of the final output layer obtained by forward transmission of the information are calculatedtruRoot mean square error E ofLAs shown in formula (5):
Figure BDA0003225465580000083
error signal ELAs a basis for modifying its weight, offset, i.e. error ELThe neurons are allocated to each level as shown in the following formula (6):
Figure BDA0003225465580000091
step four, data prediction stage
Obtaining n sub-data sets D by calculation in two stagesiX of (2)iThe load data is used as the input quantity of the load prediction model in the third step, and load data to-be-predicted points (x) with n different data freshness degrees can be obtained0,y0) Predicted load value y of electric powerpre. I.e. comparing the time points when t is equal to 0, in case of different freshness of the power data informationLoad prediction relative error value Ei，EiIndicating the utilization of the ith sub-dataset DiAnd selecting a predicted value with the minimum error as a final load predicted value and outputting corresponding data freshness.

Claims (7)

1. A power grid power data prediction method based on information freshness is characterized by comprising a power data preprocessing stage, an information freshness calculation stage, a load prediction neural network training stage and a data prediction stage;
selecting a power data sample, processing the power data sample by using an elastic BP neural network, and training in the neural network to obtain a neural network model for predicting load;
step one, preprocessing of electric power data
Processing and sampling the power data samples, and primarily classifying the processed and sampled power data samples;
step two, calculating the freshness of the information
Performing final classification on the primarily classified power data samples in the step one to obtain input power data in an output prediction stage;
step three, data prediction
And (4) taking the power data samples classified in the step two as input quantity of a neural network model for load prediction, obtaining power prediction load values of load data with different data freshness to points to be predicted, selecting a predicted value with the minimum error as a final load predicted value, and outputting the corresponding data freshness.
2. The method for grid power data prediction based on information freshness as claimed in claim 1, wherein the sampling of the first step specifically comprises: sampling the power data sample in the first step once every m minutes, wherein the sampling times are set as t, and N times of sampling are total, so that the sample set is as follows: d { (x)0,y0),(x1,y1),...,(xt,yt),...,(xN,yN) X, abscissatMt is the length of time from the point to be predicted, in seconds, and the ordinate ytPoint at time t ═ 0 (x) in kwh0,y0) As the point to be predicted for the electrical load.
3. The method for predicting the grid power data based on the information freshness as claimed in claim 2, wherein the primary classification of the first step specifically comprises:
dividing the sample set D into n parts, and using symbol D for each sub-data setiIs expressed, and the length N of the sub data seti，
Figure FDA0003225465570000021
For each subset of power data DiMean clustering with k-means to form the best classification.
4. The method for predicting the power grid power data based on the information freshness as claimed in claim 3, wherein the clustering algorithm formula is as follows:
Figure FDA0003225465570000022
order (x)j,yj) Is the jth sampling result, (a)i,bi) Is the ith data centroid, then
Figure FDA0003225465570000023
Representing the Euclidean distance between the jth sampling result and the ith mass center, and comparing the distances between the sampling point and the r mass centers so as to obtain the minimum distance
Figure FDA0003225465570000024
And (6) carrying out data classification.
5. Grid power data prediction based on information freshness as claimed in claim 1Method, characterized in that the information freshness is dependent on the subdata set DiThe weighted average of r centroids to the point to be predicted where t is 0 is defined as follows:
Figure FDA0003225465570000025
wherein (a) in the formulai,bi) Representative data subset DiThe ith centroid of (a)0,b0) I.e. the point to be predicted
Will phiiAnd subdata set DiR centroids of the load values y on the ordinateiMultiplication, the result of which is XiIs represented by Xi＝{Φiy1,Φiy2,...,Φiyr}，XiIs a column vector of r rows; mixing XiEach element X in (1)jThe Min-Max normalization of the following equation (2) is performed and then used as the input for the data prediction stage
Figure FDA0003225465570000026
6. The method for predicting the power grid power data based on the information freshness as claimed in claim 1, wherein the power data samples classified in the second step are used as input quantity of a neural network model for predicting load, power predicted load values of load data with different data freshness to points to be predicted are obtained, the predicted value with the smallest error is selected as the final load predicted value, and the corresponding data freshness is output.
7. The method for predicting the power grid power data based on the information freshness as claimed in claim 1, wherein the training process of the load prediction model is specifically described as follows:
1) information forward transfer process: output value O of input layer1And input value O of output layerLAll equal to the input value, hiding the input value x of each layer network in the layertAre the weight of the network of the layer and the output value O of the network of the previous layert-1An activation output value of the product weighted sum of;
if the information is transmitted forward from the first layer hidden layer to the t layer hidden layer, the process is as shown in formula (3):
Figure FDA0003225465570000031
for the hidden layer activation function, a sigmoid function is mainly adopted, and a specific calculation formula is shown as a formula (4):
Figure FDA0003225465570000032
2) the error back propagation process: in the process of backward transmission, firstly, the output value of the final output layer obtained by forward transmission of the information is calculated, namely the load predicted value, the load predicted value and the actual value YtruRoot mean square error E ofLAs shown in formula (5):
Figure FDA0003225465570000033
error signal ELAs a basis for modifying its weight, offset, i.e. error ELThe neurons are allocated to each level as shown in the following formula (6):
Figure FDA0003225465570000034
CN202110970503.1A 2021-08-23 2021-08-23 Power grid power data prediction method based on information freshness Pending CN113657678A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110970503.1A CN113657678A (en) 2021-08-23 2021-08-23 Power grid power data prediction method based on information freshness

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110970503.1A CN113657678A (en) 2021-08-23 2021-08-23 Power grid power data prediction method based on information freshness

Publications (1)

Publication Number Publication Date
CN113657678A true CN113657678A (en) 2021-11-16

Family

ID=78481635

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110970503.1A Pending CN113657678A (en) 2021-08-23 2021-08-23 Power grid power data prediction method based on information freshness

Country Status (1)

Country Link
CN (1) CN113657678A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114626306A (en) * 2022-03-22 2022-06-14 华北电力大学 Method and system for guaranteeing freshness of regulation and control information of park distributed energy
CN114936715A (en) * 2022-06-20 2022-08-23 中北大学 Fruit freshness prediction method and system based on smell information and storage medium
CN117253095A (en) * 2023-11-16 2023-12-19 吉林大学 Image classification system and method based on biased shortest distance criterion

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114626306A (en) * 2022-03-22 2022-06-14 华北电力大学 Method and system for guaranteeing freshness of regulation and control information of park distributed energy
CN114936715A (en) * 2022-06-20 2022-08-23 中北大学 Fruit freshness prediction method and system based on smell information and storage medium
CN114936715B (en) * 2022-06-20 2024-05-14 中北大学 Fruit freshness prediction method, system and storage medium based on smell information
CN117253095A (en) * 2023-11-16 2023-12-19 吉林大学 Image classification system and method based on biased shortest distance criterion
CN117253095B (en) * 2023-11-16 2024-01-30 吉林大学 Image classification system and method based on biased shortest distance criterion

Similar Documents

Publication Publication Date Title
CN110991786B (en) 10kV static load model parameter identification method based on similar daily load curve
CN113657678A (en) Power grid power data prediction method based on information freshness
Kang et al. A weight-incorporated similarity-based clustering ensemble method based on swarm intelligence
Chen et al. Multivariate fuzzy forecasting based on fuzzy time series and automatic clustering techniques
Wang et al. A cluster-based competitive particle swarm optimizer with a sparse truncation operator for multi-objective optimization
CN109461025B (en) Electric energy substitution potential customer prediction method based on machine learning
CN111105104A (en) Short-term power load prediction method based on similar day and RBF neural network
CN108898154A (en) A kind of electric load SOM-FCM Hierarchical clustering methods
CN109255726A (en) A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology
CN111160626A (en) Power load time sequence control method based on decomposition and fusion
CN110796159A (en) Power data classification method and system based on k-means algorithm
CN110544047A (en) Bad data identification method
Cheng et al. A novel fuzzy time series forecasting method based on fuzzy logical relationships and similarity measures
CN110177112B (en) Network intrusion detection method based on double subspace sampling and confidence offset
CN111932081A (en) Method and system for evaluating running state of power information system
CN115204698A (en) Real-time analysis method for power supply stability of low-voltage transformer area
CN113886183B (en) Method for measuring and calculating occurrence time of voltage sag event
CN113884807B (en) Power distribution network fault prediction method based on random forest and multi-layer architecture clustering
CN114219228A (en) Stadium evacuation evaluation method based on EM clustering algorithm
CN113112177A (en) Transformer area line loss processing method and system based on mixed indexes
CN117056763A (en) Community discovery method based on variogram embedding
CN115545111B (en) Network intrusion detection method and system based on clustering self-adaptive mixed sampling
CN114880754B (en) BIM-based building energy consumption management method and system
CN116307111A (en) Reactive load prediction method based on K-means clustering and random forest algorithm
CN116304773A (en) Community discovery method based on graph neural network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination